Single-Camera Pose Estimation for Real-Time Fitness Motion Counting

The patent combines multiple camera arrays and processing functions into a unified system that uses a single camera with pose estimation algorithms. Instead of requiring multiple separate cameras positioned at different angles, the system merges the tracking function into one device that captures video and processes pose data through machine learning models, significantly reducing hardware complexity while maintaining tracking precision.

The patent replaces the mechanical/optical system of multiple physical cameras with a computational approach using pose estimation algorithms and Convolutional Neural Networks. The system substitutes hardware complexity with software intelligence, using machine learning models to extract motion data from standard video feeds, thereby achieving accurate tracking without specialized camera equipment.

The patent makes the system universal by designing it to work with general-purpose computing devices like smartphones, tablets, and standard computers that already have built-in cameras. The pose estimation technology adapts to various devices and environments, allowing users to perform fitness activities at home without specialized equipment. The system processes video from different camera types and adjusts accordingly, greatly improving accessibility.

The system performs automatic pose detection and motion counting without requiring user setup or calibration. The Convolutional Neural Network automatically identifies body positions and tracks movements in real-time, providing reliable feedback without human intervention. This self-service capability eliminates the need for users to configure complex hardware or understand technical parameters, making the system easy to operate.